Stephan Frank <crimsonf@...> writes:
> On 08/09/2013 10:27 AM, Eric Marsden wrote:
>>>>>>> "sf" == Stephan Frank <crimsonf@...> writes:
>> sf> do you have an idea, which of the tests might be the culprit. I have
>> sf> completely switched to amd64 apart from an old laptop, so x86 only got
>> sf> very light testing.
>> Well, here is a more curious result. When I compile, load and run the
>> tests interactively, I can't reproduce the problem. However, it
>> remains reproducible during the standard "make.sh" build, with the
>> same GC invariant lost.
>> Interactive testing:
>> bash run-sbcl.sh --lose-on-corruption --disable-debugger --no-sysinit --no-userinit
>> (require :sb-rt)
>> (compile-file "contrib/sb-gmp/gmp.lisp")
>> (load *)
>> (compile-file "contrib/sb-gmp/tests.lisp")
>> (load *)
> I have revived my trusty old laptop (the only 32bit machine I have left)
> and have successfully build SBCL 1.1.10 including all contribs. This is
> on Ubuntu 13.10 with GMP 5.0.5. So the reason is either a different GMP
> version or something in your local setup. Alas, unfortunately I can't
> reproduce the problem on my side.
I'm not sure why you need to have a 32-bit OS when you can run 32-bit
SBCL perfectly fine on a 64-bit OS.
With best regards, Stas.